changelog

ecos实时嵌入式操作系统

2003-03-03  Jonathan Larmour  <[e-mail]>

	* tests/fptest.c: Make all variables static to avoid any risk
	of collisions with symbols defined elsewhere in eCos.

2003-02-27  Jonathan Larmour  <[e-mail]>

	* include/kapidata.h: Revert change of 2001-08-23 and instead make
	it conditional on the GCC version. Also add comments explaining why
	this file has been apparently obfuscated.

2003-02-25  Nick Garnett  <[e-mail]>

	* tests/fptest.c (alarm_fn): Added CYG_TEST_PASS() call to keep
	configtool happy.

2003-02-24  Jonathan Larmour  <[e-mail]>

	* cdl/kernel.cdl: Update doc links.
	* cdl/synch.cdl: Ditto.
	* cdl/thread.cdl: Ditto.

2003-02-19  Nick Garnett  <[e-mail]>

	* tests/fptest.c: Changed to run for a constant time rather than a
	constant number of iterations, with a shorter run time for
	simulated targets.

2003-02-10  Gary Thomas  <[e-mail]>

	* include/thread.inl: Add more debug info to thread stack checking.

	* tests/clocktruth.cxx: 
	* tests/clockcnv.cxx: Use <cyg/infra/diag.h> for diag_printf()
	prototype, rather than hard coding (because it was inconsistent).

2003-02-05  Gary Thomas  <[e-mail]>

	* tests/kcache2.c (test_dcache_operation): New test to verify
	that the DATA cache syncs properly.

2003-01-31  Nick Garnett  <[e-mail]>

	* cdl/kernel.cdl: 
	* tests/fptest.c: Added this program to test interaction of FPU
	with multiple threads.

2003-01-30  Jonathan Larmour  <[e-mail]>

	* src/common/thread.cxx: Fix potential warning and overflow with
	CYGNUM_KERNEL_THREADS_DATA_MAX == 32.

2003-01-28  Jonathan Larmour  <[e-mail]>

	* src/common/kapi.cxx (cyg_thread_get_next): Be quite zealous about
	checking the validity of passed in threads in debug mode.
	(cyg_thread_get_info): Ditto.

	* cdl/thread.cdl: Correct max legal value for
	CYGNUM_KERNEL_THREADS_DATA_MAX.

2003-01-22  Jonathan Larmour  <[e-mail]>

	* doc/kernel.sgml: Document cyg_thread_info type.

2003-01-13  Dmitriy Korovkin  <[e-mail]>
2003-01-13  Jonathan Larmour  <[e-mail]>

	* include/mqueue.hxx: Allow get/put to return time out.
	* include mqueue.inl: Ditto.

2003-01-02  Gary Thomas  <[e-mail]>

	* tests/kcache2.c: New subtest for raw data cache operations.

2002-12-12  Nick Garnett  <[e-mail]>

	* src/common/kapi.cxx: 
	* include/kapi.h:
	Added function cyg_thread_get_next(), cyg_thread_find() and
	cyg_thread_get_info() to allow the current set of threads to be
	enumerated, and per-thread information to be retrieved safely.

	* doc/kernel.sgml: Documented new KAPI calls.

	* src/common/thread.cxx: Zero unique_id in thread destructor so
	that a stale thread pointer can be checked for validity.

	* include/instrmnt.h:
	Added cyg_instrument_state() to report the current state of an
	instrumentation flag.
	Moved ifdef for CYGDBG_KERNEL_INSTRUMENT_MSGS out of within FLAGS
	ifdef. We can have messages without flags.

	* src/instrmnt/meminst.cxx: 
	Added cyg_instrument_state() to report the current state of an
	instrumentation flag.
	Modified cyg_instrument_msg() in line with header and table
	changes.

	* host/instr/dump_instr.c:
	* host/instr/instrument.sh:
	* include/instrument_desc.h: 
	Added a final NULL element to the generated table in
	instrument_desc.h to mark its end. Otherwise code that does not
	have access to the table definition cannot find its end. Also
	added ifdefs to allow instrument_desc.h to be used to acquire the
	structure definition and table pointer.

2002-12-03  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx: Add tests of 'flag' synchronizers.

2002-10-28  Andrew Lunn  <[e-mail]>

	* tests/tm_basic.cxx: Include infra/diag.h and removed the 
	incorrect prototype for diag_printf

2002-10-16  Gary Thomas  <[e-mail]>

	* include/test/stackmon.h (STACKMON_PRINTF): Use #include to get
	prototype for 'diag_printf()'.

2002-10-01  Jonathan Larmour  <[e-mail]>

	* src/common/clock.cxx (add_alarm): Tweak last change to allow
	alarm order for identical alarms to be the same as it used to be.

2002-09-30  Jonathan Larmour  <[e-mail]>

	* src/common/clock.cxx (add_alarm): Fix bug resulting in alarms
	not being added at all if a lower triggered alarm already exists.
	Reported by Christoph Csebits.
	Also fix bug when alarm is entered for the same time as tail.
	These bugs only apply for CYGIMP_KERNEL_COUNTERS_SORT_LIST enabled.

	* doc/kernel.sgml: document that order of callback for alarms at
	identical times is unspecified.

2002-08-08  Nick Garnett  <[e-mail]>

	* src/sched/sched.cxx (unlock_inner): Removed initial
	assertion. This has served its purpose and with the introduction
	of routines such as unlock_reschedule() is prone to firing in
	otherwise benign circumstances.

2002-08-05  Bart Veer  <[e-mail]>

	* cdl/kernel.cdl, include/kapidata.h, include/kapi.h:
	Allow configurations with the kernel but no malloc

2002-06-05  Gary Thomas  <[e-mail]>

	* include/kapi.h: Fix prototype (to allow builds with net stack).

2002-05-24  Jesper Skov  <[e-mail]>

	* cdl/kernel.cdl: Fix typo.

2002-05-23  Jonathan Larmour  <[e-mail]>

	* include/kapi.h: Expose new cyg_thread_add_destructor and
	cyg_thread_rem_destructor APIs.

	* include/kapidata.h (struct Cyg_Destructor_Entry):
	Change CYG_ADDRWORD to more correct cyg_addrword_t.

	* include/thread.hxx (class Cyg_Thread): add_destructor and
	rem_destructor are onyl static when not per-thread.

	* include/thread.inl (add_destructor): Don't need to lock and unlock
	scheduler when destructors are per-thread.
	(rem_destructor): Ditto.

	* src/common/thread.cxx (exit): No need to lock scheduler when
	calling destructors.

	* doc/kernel.sgml: Document thread destructor API.

2002-05-23  Nick Garnett  <[e-mail]>

	* doc/kernel.sgml: Some more edits to the kernel documentation:
	changed all references to message boxes to mail boxes, clarified
	some thing in various places, fixed a few typos.

2002-05-23  Jesper Skov  <[e-mail]>

	* cdl/kernel.cdl: Cleaned up kernel tests rule. Also remove tests
	depending on the C API when its not present. And skip dhrystone
	test when debug or instrumentation is enabled.

	* tests/tm_basic.cxx: Fixed warning.

2002-05-22  Nick Garnett  <[e-mail]>

	* doc/kernel.sgml:
	Fixed up SMP documentation.
	Rewrote condtion variable documentation to make its relationship
	to mutexes more apparent.

2002-05-22  Jesper Skov  <[e-mail]>

	* tests/smp.cxx: Move inclusion of testaux.hxx to below NA
	checks.

2002-05-21  Bart Veer  <[e-mail]>

	* doc/kernel.sgml:
	Major update to the kernel documentation.

2002-05-21  Jesper Skov  <[e-mail]>

	* src/common/clock.cxx (dsr): Fix latency ifdef guards.
	* tests/tm_basic.cxx: Same.

2002-05-20  Jonathan Larmour  <[e-mail]>

	* src/sched/mlqueue.cxx: Don't force timeslice tracing on by default
	against	the user's wishes.
	* cdl/scheduler.cdl: Instead provide an option controlling it.

2002-05-09  Jonathan Larmour  <[e-mail]>

	* include/kapidata.h (CYG_HARDWARETHREAD_MEMBERS): entry_point
	should be a pointer to function not an addrword (which may be
	different).
	(cyg_reason_t): Define cyg_reason_t as an enum not an int.
	(struct cyg_mutex_t): Define mutex protocol as an enum not a cyg_uint32.
	Define locked member as a cyg_atomic, not a cyg_bool.

2002-05-09  Nick Garnett  <[e-mail]>

	* tests/timeslice.c: Modified test thread to accumulate run time
	by reading HAL clock, and to detect and record changes of CPU. The
	previous version of this code reported varying results - probably
	as a consequence of cache effects.

	* include/instrmnt.h: Fixed left-over debug edit to MLQ
	instrumentation enable code.

2002-04-29  Jesper Skov  <[e-mail]>

	* include/kapi.h: Added cyg_alarm_get_times and
	cyg_counter_multi_tick.
	* src/common/kapi.cxx: Same. Er, fix oversight.
	* doc/kernel.sgml: Doc update.

2002-04-24  Yoshinori Sato  <[e-mail]>

	* src/sync/flag.cxx: Don't set default args in func definitions.

2002-04-24  Jesper Skov  <[e-mail]>

	* tests/dhrystone.c: Moved platform max number of passes to CDL in
	HALs.

2002-04-23  Jesper Skov  <[e-mail]>

	* tests/dhrystone.c: Allow platform to define max number of passes
	via CYGPRI_KERNEL_TESTS_DHRYSTONE_PASSES.

2002-04-09  Jonathan Larmour  <[e-mail]>

	* include/kapi.h: Use NORET attributes from cyg_type.h.
	* include/sched.hxx (class Cyg_Scheduler): Ditto.

2002-04-08  Nick Garnett  <[e-mail]>

	* src/sync/mutex.cxx: Added IF_PROTOCOL_ACTIVE macro to condition
	use of the mutex counter. This means that a mutex that is not
	configured to participate in a priority inversion protocol does not
	count for operation of that protocol. 

2002-03-04  Lars Viklund <[e-mail]>

	* include/kapidata.h (CYG_SCHEDTHREAD_ASR_MEMBER): 
	asr_inhibit is cyg_ucount32 - must match

2002-02-13  Hugo Tyson  <[e-mail]>

	* doc/kernel.sgml: NEW FILE: Correct reference to uITRON doc.

2002-01-28  Jesper Skov  <[e-mail]>

	* tests/kexcept1.c: Fix warning.

2002-01-24  Jesper Skov  <[e-mail]>

	* tests/kcache1.c (time0DI): Disable interrupts around the
	sync+invalidate process.
	* tests/kcache2.c: Same.

2002-01-23  Jonathan Larmour  <[e-mail]>

	* src/intr/intr.cxx (mask_interrupt): Disable interrupts.
	(unmask_interrupt): Ditto.
	(mask_interrupt_intunsafe): New function to avoid disabled interrupts.
	(unmask_interrupt_intunsafe): Ditto.
	* include/intr.hxx (class Cyg_Interrupt): Declare above new funcs.
	* src/common/kapi.cxx (cyg_interrupt_unmask_intunsafe): Above
	replicated to KAPI.
	(cyg_interrupt_mask_intunsafe): Ditto.
	* include/kapi.h: Ditto.

2002-01-07  Nick Garnett  <[e-mail]>

	* tests/except1.cxx: 
	* tests/kexcept1.cxx:
	Added ifdef to disable these tests in ARM PID platform. This
	platform cannot generate any of the exceptions that this test
	needs to work.

2001-12-12  Jesper Skov  <[e-mail]>

	* tests/kcache2.c (time_ilock): Put some dummy goto statements in
	to prevent compiler from moving labels around.

2001-11-29  Jonathan Larmour  <[e-mail]>

	* src/sched/mlqueue.cxx (timeslice_cpu): Reset timeslice_count on
	a timeslice.
	Noticed by Tony Kho.

2001-11-23  Jonathan Larmour  <[e-mail]>

	* tests/klock.c (entry1): Support running with
	CYGFUN_KERNEL_THREADS_TIMER disabled.

2001-10-30  Nick Garnett  <[e-mail]>

	* tests/kcache2.c (test_dsync):
	Added call to HAL_ICACHE_INVALIDATE_ALL() just before DCACHE
	disable. In platforms where both caches are controlled together,
	such as the VR4300, not doing this can result in strange behaviour
	as the ICACHE gets turned off under the program's feet, when it is
	not ready for it.

2001-10-17  Jesper Skov  <[e-mail]>

	* tests/dhrystone.c: CYGINT_ISO_STRING_STRFUNCS check changed to
	ifdef.

2001-10-12  Jonathan Larmour  <[e-mail]>

	* cdl/synch.cdl (CYGIMP_KERNEL_SYNCH_MQUEUE_NOT_INLINE): New option
	to avoid inlining mqueue implementation.

	* src/sync/mqueue.cxx: New file to provide non-inline version of
	mqueue functions.

	* include/mqueue.hxx: Don't include .inl if user doesn't want inlining.

	* include/mqueue.inl: Allow choice of inlining or not to be
	overridden by macro.

	* tests/mqueue1.cxx: Never use inline version.

2001-10-11  Jesper Skov  <[e-mail]>

	* tests/kmutex3.c (new_thread): Fixed allocation: increase counter
	before starting threads which have been allocated resources.
	* tests/kmutex4.c (new_thread): Same.
	* tests/mutex3.cxx (new_thread): Same.
	* tests/testaux.hxx (new_thread): Same.

	* tests/kcache2.c: Fixed warning.
	* tests/stress_threads.c: Same.

2001-09-20  Jonathan Larmour  <[e-mail]>

	* host/instr/dump_instr.c (main): Fix argc check.

2001-09-07  Jonathan Larmour  <[e-mail]>

	* include/thread.inl (measure_stack_usage): Deal with stack limits.

2001-09-07  Nick Garnett  <[e-mail]>

	* src/common/thread.cxx (set_priority): Change argument to
	instrumentation call to new_priority rather than current
	priority. (Suggested by Andrew Lunn).

2001-09-05  Jesper Skov  <[e-mail]>

	* include/mqueue.inl (Cyg_Mqueue): Initialize busy flag of last
	entry in the list.